AI Agent Hub
Back to plugins
dsh-plugin-web-search-tavily preview

dsh-plugin-web-search-tavily

Web Tools Updated 2026.08.15

Run the following command in DeepSeek Harness:

dsh plugin install coderdailyone/dsh-plugin-web-search-tavily

Paste the following prompt into your AI chat to install this plugin:

Run dsh plugin install coderdailyone/dsh-plugin-web-search-tavily inside DeepSeek Harness (source: https://github.com/coderdailyone/dsh-plugin-web-search-tavily); the plugin self-registers as the Tavily backend for the web_search tool.

About this plugin

Inside DeepSeek Harness the web_search tool is the single entry point for web retrieval in the model loop, yet it ships without a fixed search engine behind it. dsh-plugin-web-search-tavily fills that gap by registering a Tavily-backed WebSearchProvider on the ctx.web capability seam, so every web_search call is silently served by Tavily while the model-facing schema and rendering stay completely untouched.

Under the hood the plugin maps Tavily titles, snippets, publication dates, and generated answers into the tool-layer fields the model already understands: snippet, title, publishedAt, and content. You can toggle between basic and advanced search depth, switch the topic vertical between general and news, and decide whether to include the one-paragraph answer Tavily generates. Cancellations surface as WEB_ABORTED; every other failure is classified as WEB_PROVIDER_ERROR with the original cause chain preserved for easy debugging. The availability check is purely local and never hits the network.

This plugin is for dsh developers who want Tavily as their search backend to get structured results, news timelines, or AI-generated summary answers, without writing a single line of model-facing code. When multiple providers are present a single env var pins the choice to Tavily; when Tavily is the only option it activates automatically, making the setup genuinely zero-config.

Screenshots

Use Cases

  • Swap the dsh web_search backend to Tavily for structured snippets and AI-generated answers
  • Query the news vertical to get publication dates for time-sensitive aggregation pipelines
  • Enable advanced depth plus generated answers to feed richer candidates into a downstream RAG flow

Best For

  • dsh developers who have web_search enabled and want to change the search backend
  • Engineers wiring high-quality web retrieval into multi-turn conversational pipelines
  • Application authors who need news timelines or AI summaries instead of raw keyword results